Overriding commands for target Android Makefile - android-ndk

I am trying to compile one of the modules in my Android ndk project with g++ although the sources are all in C.
My eyes are irritated by the make system warnings:
`C:/NVPACK/android-ndk-r8d/build/core/build-binary.mk:348: warning: overriding commands for target 'obj/local/armeabi/objs/xxx/yyy.o'`
`C:/NVPACK/android-ndk-r8d/build/core/build-binary.mk:345: warning: ignoring old commands for target 'obj/local/armeabi/objs/xxx/yyy.o'`
And these warning pairs will be printed as much as there will be the source files and therefore the objects.
I've tried to declare LOCAL_SRC_FILES with all the different flavors.
`LOCAL_SRC_FILES :=
$(LOCAL_PATH)/Directory/source.c
$(notdir $(wildcard $(LOCAL_PATH)/*.c))
$(notdir $(wildcard $(LOCAL_PATH)/Directory/*.c))
$(addprefix DirectoryPrefix/,$(notdir $(wildcard $(LOCAL_PATH)/Directory/*.c)))`
And still the warning persists.
Make document says:
warning: overriding commands for target xxx''
warning: ignoring old commands for target xxx''
GNU make allows commands to be specified only once per target (except for double-colon rules). If you give commands for a target which already has been defined to have commands, this warning is issued and the second set of commands will overwrite the first set.
But I cannot understand how this is related at all.
After dealing with it seems like making a g++ compile theses C files makes this warning appear.
Therefore specifying this statement:
LOCAL_CPP_EXTENSION := .c
Which makes C files build with g++ is causing it. Because when compiling with gcc no warnings are printed.

Make sure that you have included clear vars:
include $(CLEAR_VARS)
And if building any other libraries that you include the correct build macro like
include $(BUILD_SHARED_LIBRARY)
I had run into this a day or two ago and it was while adding a new library I had forgotten to include one or the other (I think the CLEAR_VARS makes most sense.) So it was re-appending some values from the main library to the child library or vice versa.

Related

How can get g++ to use my own glibc build's headers correctly?

There's a TL;DR at the end if the context is too much!
Context
I am trying to update the version of glibc a project uses to 2.23 (I know it's old, that's another issue). In order to do this, I need to swap out the libraries and use the associated interpreter.
I encountered some issues when swapping out the interpreter that looked like an ABI change, so I figured it was probably because the header files had changed somehow and started working on getting those included into the project.
At first I tried using -I to include the headers, but got an error (see below). Later I tried setting --sysroot, but this quickly felt like the wrong way of doing things since I was essentially reinventing what g++ already did with system headers. I later found another mechanism that looked more promising (see Problem section).
Could this be an XY issue? Absolutely, but either way, the problem I'm seeing seems odd to me.
Problem
I looked into whether there was a different mechanism to include headers for system libraries, such as glibc, in gcc and g++. I found the flag -isystem:
-isystem dir
Search dir for header files, after all directories specified by -I but before the standard system directories. Mark it as a system directory, so that it gets the same special treatment as is applied to the standard system directories. If dir begins with "=", then the "="
will be replaced by the sysroot prefix; see --sysroot and -isysroot.
I figured that this was probably wanted and set about intergrating this flag into the build system for the project. The resulting g++ command looks like this (simplified and broken onto multiple lines):
> /path/to/gcc-6.3.0/bin/g++
-c
-Wl,--dynamic-linker=/path/to/glibc-2.23/build/install/lib/ld-linux-x86-64.so.2
-Wl,--rpath=/path/to/glibc-2.23/build/install/lib
-isystem /path/to/glibc-2.23/build/install/include
-I.
-I/project-foo/include
-I/project-bar/include
-o example.o
example.cpp
This leads to the following error, followed by many similar ones:
In file included from /usr/include/math.h:71:0,
from /path/to/gcc-6.3.0/include/c++/6.3.0/cmath:45,
from example.cpp:42:
/path/to/glibc-2.23/build/install/include/bits/mathcalls.h:63:16: error: expected constructor, destructor, or type conversion before '(' token
__MATHCALL_VEC (cos,, (_Mdouble_ __x));
Looking into this, it appears that this particular math.h is incompatible with this version of glibc. The fact it tries to use it surprises me, because the math.h file exists in the glibc directory I specified; why didn't it use that? Here's how I verified that file exists:
> ls /path/to/glibc-2.23/build/install/include/math.h
/path/to/glibc-2.23/build/install/include/math.h
Research
I searched around on the internet for people with a similar issue and came across the following relevant things:
https://github.com/riscv/riscv-gnu-toolchain/issues/105
https://askubuntu.com/questions/806220/building-ucb-logo-6-errors-in-mathcalls-h
-isystem on a system include directory causes errors
The last of these is the most promising; it talks about why -isystem won't work here stating that the special #include_next traverses the include path in a different way. Here, the solution appears to be "don't use -isystem where you can help it", but since I've tried using -I only get get the same problem again, I'm not sure how I'd apply that here.
Original issue
When compiling with the new glibc, I get the following error (our build process ends up running some of the programs it compiles to generate further source to be compiled, hence this runtime error whilst compiling):
Inconsistency detected by ld.so: get-dynamic-info.h: 143: elf_get_dynamic_info: Assertion `info[DT_RPATH] == NULL' failed!
I found a couple of relevant things about this:
https://www.linuxquestions.org/questions/linux-software-2/how-to-get-local-gcc-to-link-with-local-glibc-404087/
https://www.linuxquestions.org/questions/programming-9/inconsistency-detected-by-ld-so-dynamic-link-h-62-elf_get_dynamic_info-assertion-621701/
The only solution I see there is completely recompiling gcc to use the new glibc. I'd like to avoid that if possible, which is what lead me down the include route.
Eliminating the complex build system
To try and eliminate the complex build system on the "real" project, I reproduced the problem using the following test.cpp file:
#include <cmath>
int main() {
}
Compiled using:
> /path/to/gcc-6.3.0/bin/g++ test.cpp -Wl,--dynamic-linker=/path/to/glibc-2.23/build/install/lib/ld-linux-x86-64.so.2 -Wl,--rpath=/path/to/glibc-2.23/build/install/lib
Running yields the same original issue:
> ./a.out
Inconsistency detected by ld.so: get-dynamic-info.h: 143: elf_get_dynamic_info: Assertion `info[DT_RPATH] == NULL' failed!
Trying to use the newer headers yields the same include issue:
> /path/to/gcc-6.3.0/bin/g++ test.cpp -Wl,--dynamic-linker=/path/to/glibc-2.23/build/install/lib/ld-linux-x86-64.so.2 -Wl,--rpath=/path/to/glibc-2.23/build/install/lib -isystem /path/to/glibc-2.23/build/install/include
In file included from /usr/include/math.h:71:0,
from /path/to/gcc-6.3.0/include/c++/6.3.0/cmath:45,
from test.cpp:1:
/path/to/glibc-2.23/build/install/include/bits/mathcalls.h:63:16: error: expected constructor, destructor, or type conversion before '(' token
__MATHCALL_VEC (cos,, (_Mdouble_ __x));
TL;DR
How can I get g++ to include the headers from my glibc build correctly, without it accidentally including incompatible files from /usr/include?
In your GCC version,<cmath> uses #include_next, which means that you need to make sure that the directory which contains the cmath file comes before (on the include search path) the directory with the proper math.h for the version of glibc you are building against.
You can use g++ -v to view the search path. In your case, it probably looks like this:
#include "..." search starts here:
#include <...> search starts here:
.
/project-foo/include
/project-bar/include
/path/to/glibc-2.23/build/install/include
/usr/include/c++/6
/usr/include/x86_64-linux-gnu/c++/6
/usr/lib/gcc/x86_64-linux-gnu/6/include
/usr/local/include
/usr/lib/gcc/x86_64-linux-gnu/6/include-fixed
/usr/include/x86_64-linux-gnu
/usr/include
If you configure glibc with --prefix=/usr and install it with DESTDIR=/path/to/glibc-2.23/build/install, its header files will be installed into the directory /path/to/glibc-2.23/build/install/usr/include. This means you should be able to use the -isysroot option, which rewrites the default /usr/include directory, resulting in the right ordering of the search path:
#include "..." search starts here:
#include <...> search starts here:
.
/project-foo/include
/project-bar/include
/usr/include/c++/6
/usr/include/x86_64-linux-gnu/c++/6
/usr/include/c++/6/backward
/usr/lib/gcc/x86_64-linux-gnu/6/include
/usr/lib/gcc/x86_64-linux-gnu/6/include-fixed
/path/to/glibc-2.23/build/install/usr/include

Why does CMake not respect LIBRARY_PATH and CPATH

I'm using CMake on a HPC system that uses modules. Those modules usually set LIBRARY_PATH and CPATH so one can simply include headers and link against libraries without additional -L or -I.
However when using CMake those libraries must be found by CMake. I would expect, that LIBRARY_PATH and CPATH were automatically considered by CMake and captured in e.g. CMAKE_SYSTEM_LIBRARY_PATH and CMAKE_SYSTEM_INCLUDE_PATH but they are not.
Is there any reasoning why this was (not) done?
Wouldn't it be a good idea (for now manually, later automatically by CMake) to append LIBRARY_PATH to CMAKE_SYSTEM_LIBRARY_PATH and CPATH to CMAKE_SYSTEM_INCLUDE_PATH?
Is there any reasoning why this was (not) done?
Wouldn't it be a good idea (for now manually, later automatically by CMake) to append LIBRARY_PATH to CMAKE_SYSTEM_LIBRARY_PATH and CPATH to CMAKE_SYSTEM_INCLUDE_PATH?
There is another view that having your build depend on environment variables results in unreliable builds. From GNU Make manual:
It is not wise for makefiles to depend for their functioning on environment variables set up outside their control, since this would cause different users to get different results from the same makefile. This is against the whole purpose of most makefiles.

Makefile works in Linux but not Windows - can't find header files

Reasonably straight forward makefile. I'm trying to build an .so and a .dll to link to another program and the Linux version has gone completely smoothly. However, I receive a strange error when compiling in Windows
make: *** No rule to make target '/L/folder/folder/include/header.h',
needed by 'obj/file.o'. Stop.
Which implies it can't find "header.h" for some reason because it tries to build it.
I have this bit of code to determine the platform (ugly, I know)
ifeq ($(PLT), lnx32)
INC_DIR = /mnt/L/folder/folder/include/
TARGET = lnx32/libmine.so
endif
ifeq ($(PLT), win32)
INC_DIR = /L/folder/folder/include/
TARGET = win32/libmine.dll
endif
Some even uglier makefile code to compile file.o
$(OBJ_DIR)/file.o: $(INC_DIR)/header.h $(INC_DIR)/another_header.h $(SRC_DIR)/file.c
$(CC) $(CFLAGS) $(CPPFLAGS) $(SYS5FLAG) -c $(SRC_DIR)/file.c -o $# -I$(INC_DIR)
Not sure if it's relevant, but L is a network drive, I'm using Debian through VNC and msys in Windows. Does anything have an idea what could be going on? Thanks in advance.
UPDATE:
I changed
INC_DIR = /L/folder/folder/include/
to
INC_DIR = L:\produkter\RD\shared\include
to see what would happen and it gives a ton of
Cannot export y_varde: symbol not found
./obj/file.o: In function `some_function':
/mnt/L/folder/folder/./src/file.c:191: undefined reference to `vsprintf'
errors. It's strange that it's looking at /mnt/ though as it really shouldn't be (in Windows).
** UPDATE 2 **
Okay, the problem was due to old object files from the Linux build. Instead of rebuilding them for Windows, it assumed they were the "correct" ones and tried to link them which obviously won't work. Still have an error though, but it seems more manageable, something to do with paths...
l:\foloder\folder/./src/somefile.c:1821: undefined reference to `debug'

QMAKE_CXXFLAGS doesn't work

I'm reading everywhere that to add flags to the compiler we have to, inside the .pro, add command like this:
QMAKE_CXXFLAGS += -Wimplicit-function-declaration
Right, I'm under Linux 64bit, Qt5 and adding this inside my .pro doesn't change anything
It turns out that I don't see the flag inside the compilation line, and the CXXFLAGS inside the Makefile generated does not contain my flag !?
I found that sometimes qmake is a bit lazy and won't regenerate the Makefile even though the .pro has changed. Try rebuilding everything, or at least deleting the Makefile.

Driver/module cross compilation

I am trying to cross compile a driver for an arm based board. In the make file the search path for include files is that of the host machine's kernel i.e it points to the linux headers that come with ubuntu.
I also have the kernel source tree of target board present on the host system(i7/ubuntu).
My Question is that which include path is needed for cross compilation (native system's linux headers path or the path to the kernel source tree of board ?
Does same thing applies to all modules including drivers?
Here is a Makefile for an out of tree driver. The architecture, toolchain and kernel dir specified :
ifneq ($(KERNELRELEASE),)
# We were called by kbuild
obj-m += fpgacam.o
else # We were called from command line
KDIR := path/to/your/target/kernel
PWD := $(shell pwd)
CROSS=arm-none-linux-gnueabi-
default:
#echo ' Building Cam drivers for 2.6 kernel.'
#echo ' PLEASE IGNORE THE "Overriding SUBDIRS" WARNING'
$(MAKE) -C $(KDIR) SUBDIRS=$(PWD) ARCH=arm CROSS_COMPILE=$(CROSS) modules
install:
./do_install.sh *.ko
endif # End kbuild check
######################### Version independent targets ##########################
clean:
rm -f -r *.o *.ko .*cmd .tmp* core *.i
When make is called from the module directory, the command line path is taken, and make is redirected to the kernel directory build system using make -C. The kernel build system then the different variable passed to it to go back into the module directory with everything setup (include path, toolchain etc ..) to compile a module. This second time through the Makefile, the kbuild path is taken, and the module is built as if it was in-tree.
Compiling with your build-system's headers is spectacularly bad news, and may subtle binary incompatibilities that manifest themselves as seemingly bizarre crashes on the target.
As you've already discovered, the kernel is already hardened against this and will refuse to load modules built against the wrong headers. You'll need to build using the same source tree as the existing kernel - including any patches. You might as well at this point rebuild the entire kernel.
The kernel tree is self-contained, so simply cross-compiling it in place will work. If you're adding a driver it's probably easiest to compile it in-tree.
If you want to build any user-space components, you have two solutions:
Pass the --sysroot=<dir> option to gcc, where <dir> is the system root of your target system
Build gcc and binutils to use as their default sysroot
The latter approach is the one that Angstrom uses, and it save a lot of butt-hurt.
You might want to try using crosstool-ng
It takes care of the majority of the work. You only need to bother about the configuration settings you want to enable for kernel compilation.
Here another reference: link

Resources